Re: [patches] Cross-testing changes vs. native testing

[Jim Blandy <jimb@xxxxxxxxxxxxxxxx>]

"Joseph S. Myers" <joseph@xxxxxxxxxxxxxxxx> writes:
> On Mon, 29 Jan 2007, Jim Blandy wrote:
>
>> Joseph, in our CodeSourcery meeting this morning, you mentioned that I
>> should be sure that my changes for cross-testing don't break native
>> testing.  Empirically, I will be comparing before-and-after native
>> results, but analytically, did you notice anything that looked odd in
>> my patch?  Are there specific areas you anticipate will require extra
>> care?
>
> Just make sure that the cases where an extra argument is passed to shell 
> scripts as a cross testing wrapper are handled correctly.  If that 
> argument will be empty, that means ensuring it is quoted on the shell 
> script command line in the makefiles, but not quoted when actually used 
> within the scripts themselves.  (The cases I spot checked looked
> OK.)

As it turns out, the wrapper values I've been using have always
contained embedded spaces, so if I've forgotten to quote them, I've
encountered an error immediately.  So there shouldn't be errors of
that sort.

I did run into a case where arguments to a shell scripts are sometimes
(deliberately) null, so the caller gets no $@ entry for them at all;
passing the wrapper at the end of that argument list didn't work
well. :)

> Plain "make check" should work with no test failures on i686-pc-linux-gnu 
> native (with a suitable compiler) so provides a reasonable sanity check 
> that this is working.

All right.